Childhood Brain Tumor Study Uncovers KDM2B as a Selective Epigenetic Vulnerability

Medulloblastoma, the most common malignant brain tumor of childhood, has long been known to be a disease of many faces. Clinicians and researchers divide it into four molecular subgroups—WNT, SHH, Group 3 and Group 4—each with distinct demographics, biology and prognosis. Yet despite decades of genomic scrutiny, one question has remained stubbornly open: how do the frequent alterations in chromatin-modifying genes, found in roughly half of all cases, actually translate into the transcriptional programs that drive malignancy? A large international study now offers a strikingly concrete answer, mapping the chromatin landscape of medulloblastoma in unprecedented detail and identifying a single epigenetic regulator, KDM2B, as a dependency that is selectively essential for the two highest-risk subgroups.

The research, led by teams at St. Jude Children’s Research Hospital in Memphis and the Hopp Children’s Cancer Center in Heidelberg, took a deliberately multi-modal approach. Rather than relying on DNA sequencing alone, the investigators profiled six key histone post-translational modifications across 52 fresh-frozen tumor specimens representing all four consensus subgroups. These modifications—H3K4me1, H3K4me3, H3K27me3, H3K27ac, H3K9me3 and H3K36me3—were assayed by chromatin immunoprecipitation followed by sequencing, generating 254 ChIP-seq experiments, of which 229 were newly produced for the study. The histone data were then integrated with whole-genome sequencing, DNA methylation profiling and RNA sequencing datasets covering hundreds of additional tumors, creating one of the most comprehensive epigenomic portraits of any pediatric cancer to date.

To make sense of this torrent of data, the team applied ChromHMM, a hidden Markov model-based algorithm, using the Roadmap 18-state segmentation framework to annotate the genome of each tumor into combinatorial chromatin states. The results were immediately reassuring: clustering tumors by their chromatin states largely recapitulated known subgroup identity, and active chromatin states accurately predicted well-established subgroup signature genes. The ATOH1 and GLI2 loci, hallmarks of proliferating granule neuron progenitors, carried active chromatin specifically in SHH tumors, while the HLX promoter was active only in Group 3 and EOMES, a marker of unipolar brush cell-fated progenitors in the rhombic lip, was active in both Group 3 and Group 4. In other words, subgroup identity is not merely correlated with transcription—it is epigenetically encoded.

But the most consequential finding emerged when the researchers compared the relative abundance of individual chromatin states across subgroups. One state stood out dramatically: the bivalent enhancer state, or EnhBiv, defined by the coexistence of the enhancer mark H3K4me1 and the Polycomb mark H3K27me3. EnhBiv marks regulatory elements that are poised for activation—silent but ready to fire during development. Its genomic coverage was highest in Group 4, followed by Group 3, and lowest in WNT and SHH tumors. Intriguingly, while the classical bivalent promoter state, TssBiv, was distributed evenly across subgroups, EnhBiv was specifically enriched at promoter-proximal regions in Groups 3 and 4, occupying the flanks of transcription start sites rather than their cores, with distinct DNA methylation, GC content and CpG density profiles that set it apart from canonical poised promoters.

The genes marked by this state told a coherent story. The team identified 666 promoters significantly enriched for EnhBiv in Group 4 tumors, which they designated the Group 4 EnhBiv signature. These genes were expressed at their lowest levels precisely in Group 4, and pathway analysis revealed a striking enrichment for neurogenesis and transcriptional regulation terms—exactly the programs one would expect to be held in developmental checkmate by bivalent chromatin. Genes such as PAX3 and PAX2, regulators of neuronal differentiation, were repressed in Groups 3 and 4 through context-dependent mechanisms, with the PAX3 promoter bearing EnhBiv in Group 4 but a fully repressive Polycomb state in Group 3. The implication was clear: Group 3 and 4 tumors, which share a developmental origin in the rhombic lip subventricular zone, are holding neuronal differentiation genes in a poised, repressed state.

The hunt for the regulator responsible led to a single dominant candidate. Enrichment analysis using publicly available chromatin occupancy datasets identified KDM2B, a histone lysine demethylase and core component of the PRC1.1 Polycomb complex, as the most enriched regulator at EnhBiv-marked promoters. KDM2B was also the most significantly upregulated candidate in Group 4 tumors, the only differentially expressed enzyme among the H3K4 and H3K27 methyltransferases and demethylases, and its own promoter carried active chromatin marks in Groups 3 and 4. Elevated KDM2B expression correlated with somatic alterations in other chromatin-modifying genes, including PRDM6, KDM6A and KMT2C, hinting at layered epigenetic dysregulation. Notably, KDM2B expression is normally highest early in cerebellar development and declines during maturation, but its levels in the unipolar brush cell lineage do not fully explain the tumor-specific elevation—suggesting the tumors are actively maintaining an embryonic chromatin program.

Direct validation followed. Using CUT&RUN, a technique that maps protein-DNA interactions in intact chromatin, the researchers examined KDM2B occupancy in patient-derived xenografts representing SHH, Group 3 and Group 4 medulloblastoma. Group 3 and 4 EnhBiv-enriched promoters showed significantly higher KDM2B binding than SHH tumors, with the strongest signal in Group 4, accompanied by elevated H3K27me3. The functional consequences were equally decisive. CRISPR-mediated knockout of KDM2B in three Group 3 cell lines—D425, D283 and CHLA-01-MED—robustly suppressed growth in competition assays, while SHH medulloblastoma and high-grade glioma models were unaffected. Because authentic Group 4 culture models do not exist, the team turned to two Group 4 xenografts, DMB006 and Med2312FH, and showed in vivo that KDM2B loss impaired tumor growth in mouse cerebellum. KDM2B, in short, is a selective dependency of the high-risk subgroups.

To rule out long-term adaptation artifacts and probe therapeutic potential, the researchers engineered a degradation tag into the endogenous KDM2B alleles of D283 cells. Treatment with dTAG compounds degraded the KDM2B protein within two to four hours, and acute loss of the protein reduced proliferation by roughly twenty percent and slightly increased apoptosis in vitro. In a flank xenograft model, seven weeks of dTAG treatment reduced tumor growth and significantly prolonged survival, with immunohistochemistry confirming reduced proliferation. Transcriptomic analysis revealed what KDM2B was actually doing: its degradation predominantly upregulated gene expression, with PRC2 targets, bivalent-marked genes and neuronal differentiation programs rising to the top of the enrichment lists. At the chromatin level, KDM2B loss stripped H3K27me3 from the promoters of upregulated genes and increased H3K4me3, collapsing the EnhBiv state—while global histone modification levels, measured by mass spectrometry, remained essentially unchanged, indicating locus-specific rather than genome-wide effects.

Mechanistically, the study resolved the order of operations in exquisite temporal detail. KDM2B is known to recruit the PRC1.1 complex to unmethylated CpG islands through its CXXC DNA-binding domain, where PRC1.1 deposits H2AK119 ubiquitination, which in turn instructs PRC2 to deposit H3K27me3. Co-immunoprecipitation confirmed that KDM2B physically interacts with PRC1.1 components BCOR, PCGF1 and RYBP, and more weakly with PRC2 components including EZH2, SUZ12 and JARID2. After acute KDM2B degradation, KDM2B and BCOR were lost from repressed promoters within twelve hours, H2AK119ub and EZH2 occupancy declined in parallel, and H3K27me3 followed. Rescue experiments with domain mutants sealed the argument: wild-type KDM2B and JmjC demethylase-defective mutants fully rescued the growth defect, whereas a mutant lacking the CXXC domain failed entirely. KDM2B sustains malignancy through its DNA-binding scaffolding function, not its catalytic demethylase activity.

The therapeutic implications are profound and, unusually for an epigenetic target, well-defined. Because KDM2B’s demethylase activity is dispensable for tumor maintenance, classical catalytic inhibitors are unlikely to help; instead, the dTAG results point toward degrader technologies—proteolysis-targeting chimeras or molecular glues—that physically eliminate the protein. The study also found that multiple PRC1.1 and PRC2 subunits are highly expressed in Groups 3 and 4 and represent Group 3-selective dependencies in the DepMap CRISPR screen, supporting evaluation of existing EZH2 inhibitors and Polycomb-directed approaches as alternatives. With no effective targeted therapies currently available for Group 3 and 4 medulloblastoma, and with the KDM2B-associated EnhBiv state predicted to contribute to malignancy across a broad fraction of patients, the KDM2B–PRC1.1–PRC2 axis now stands as one of the most compelling candidate vulnerabilities in pediatric brain tumor biology—and the multidimensional datasets generated along the way, freely accessible through an online portal, promise to fuel the next wave of discovery.

Subject of Research: Epigenomic regulation of medulloblastoma subgroups and the role of KDM2B as a selective dependency

Article Title: Divergent medulloblastoma chromatin states disclose KDM2B as a selective dependency

Article References: Tao, R., Erkek-Ozhan, S., Xu, B., Matsui, Y., Mittal, P., Smith, K. S., Li, Y., Filipovic, D., Xu, R., Gao, Q., Darrow, E., Kumar, R., Djekidel, N., Bajpai, R., Hadley, J., Batts, M., Lewis, S. A., Soliman, T., Reilly, C., … Northcott, P. A. (2026). Divergent medulloblastoma chromatin states disclose KDM2B as a selective dependency. Nature Genetics, 58(10), 2590-2604. https://doi.org/10.1038/s41588-026-02745-1

Image Credits: AI Generated

DOI: 10.1038/s41588-026-02745-1

Keywords: medulloblastoma, KDM2B, epigenomics, chromatin states, Polycomb, PRC1, PRC2, bivalent enhancer, pediatric brain tumor, H3K27me3, targeted protein degradation, neuronal differentiation